DKD-R - 3-9 - 2018 (1) English

Machine Translated by Google
Physically
Technical
federal institution
policy Continuous calibration of

DKD-R 3-9 force transducers according to
the comparison method
Edition 09/2018
https://doi.org/10.7795/550.20180823J
DKD-R 3-9
Continuous calibration of force transducers using the 09/2018
Output:
comparison method https://doi.org/
Revision: 0
10.7795/10.7795/550.20180823J
Side: 2 / 65
German Calibration Service (DKD)
Calibration laboratories from industrial companies, research institutes, technical

authorities, monitoring and testing institutions have been united in the DKD since it was
founded in 1977. On May 3rd, 2011, the DKD was re-established as a technical body of the
PTB and the accredited laboratories.
This body is called the German Calibration Service (DKD) and is under the direction of the
PTB. The guidelines and guidelines developed by the DKD represent the state of the art in
the respective technical field and are available to the German Accreditation Body GmbH
(DAkkS) for the accreditation of calibration laboratories.
The accredited calibration laboratories are accredited and monitored by the DAkkS as the
legal successor to the DKD. They carry out calibrations of measuring devices and material
measures for the measured variables and measuring ranges specified during accreditation.
The calibration certificates they issue are proof of traceability to national standards, as
required by the DIN EN ISO 9000 family of standards and DIN EN ISO/IEC 17025.
Contact:
Physical-Technical Federal Institute (PTB)
DKD office
Federal Avenue 100 38116 Braunschweig
PO Box 33 45 38023 Braunschweig
Telephone secretariat: (05 31) 5 92-8021
Internet: www.dkd.eu
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 3 / 65
Suggested citation for citing the

source: Guideline DKD-R 3-9 Continuous calibration of force transducers using the
comparison method, issue 09/2018, revision 0, Physikalisch-Technische Bundesanstalt,
Braunschweig and Berlin. DOI: https://doi.org/10.7795/550.20180823J
The work including all its parts is protected by copyright and is subject to the Creative
Commons user license CC by-nc-nd 3.0 (http://creativecommons.org/licenses/by-nc nd/3.0/
de/). In this context, non-commercial (NC) means that the work may not be distributed or
made publicly available for revenue generation. Use of its content for commercial use in
laboratories is expressly permitted.
Authors:
Members of the Technical Committee for Mechanical Variables (later Force and
Acceleration) of the DKD in the period from 2005 to 2009.
Published by the Physikalisch-Technische Bundesanstalt (PTB) for the German Calibration

Service (DKD) as a result of the cooperation between the PTB and the Technical Committee
for Force and Acceleration of the DKD.
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 4 / 65
Table of Contents
foreword 6
Purpose and Scope 6
symbols and designations 7
calibration item 8
1 calibration capability 8
23 Clamping parts, built-in parts 9
4 Signal-carrying components in the K-BNME 9
4.1 Environmental conditions 9
4.2 Performing the calibration 10
4.3 Preparations for calibration 10
5 6 6.1 6.1.1 Determination of design-specific measurement uncertainty 10
contributions 6.1.2 Temperature compensation 6.2 Calibration process 6.2.1 10

Preloads 6.2.2
for shortened
Time management
calibration
based
processes
on DIN6.2.4
EN ISO
Recording
376 6.2.3
of Time
Measured
management
value 10
evaluation 7 8 Measurement uncertainty 8.1 Influencing variables 8.2 Model of the 10
evaluation "Transmission coefficient" model 8.2.1 8.2.2 "Force" model (polynomial 10
coefficients of a 3rd degree adjustment, DIN EN ISO 376) 11

12
12
15
15
17
17
22
8.3 Visualization of the measurement uncertainty budget 26
8.4 Uncertainty interval Single value indication as 26
8.5 calibration result Visualization of the calibration 26
8.6 result 28
8.6.1 Deviation and measurement uncertainty 29
8.6.2 Entry value information 9 Specifications in 30
the calibration certificate
laboratory 10 11Records in the
Conformity 12calibration 32
rules 13.2 Calibration
Literature mark Literature 13 13.1 Technical 33
33
33
34
34
35
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 5 / 65
Annex A 37
A.1 Force BNME traceability A.1.1 General 37
A.2 A.3 37
Requirements for reference and transfer force transducers Proof of 40
suitability of the transfer standards before the first use of continuous operation (proposal)
41 A.3.1 Determination of the influence of the load speed on the measured values 41 A.3.2
Determination of the influence
Determination of theof the loadof
influence speed on the
the force hysteresis parts
introduction 42 A.3.3 Test of
44 A.3.5 creep behavior
Determination 43 A.3.4
of the
influence of temperature 44 A.3.6 Recommended limit values for time-dependent parameters and temperature
behavior 45 A.3.7 Monitoring of continuous operation 45 Determination of the smallest specifiable
measurement uncertainty of a K-BNME with Reference force transducer system for continuous operation
(suggestion)
A.4
45
A.4.1 Influencing variables 45
A.4.2 Model A.4.2.1 46
Uncertainty analysis A.4.2.2 49
Measurement uncertainty budget A.4.3 50
Correction factor Kconnection, instead of the static connection measurement with transfer force
transducers 51
A.4.3.1 Uncertainty analysis 53 A.4.3.2 Measurement uncertainty budget 54 A.4.4 Correction factor Kver of
the verification of the continuous method 55 A.4.4.1 Uncertainty analysis 57 Appendix B 59 B.1 Determination
of
an optimal polynomial degree to reduce the influence of random deviations 59 B.2 Example of the effect of
the approximation with polynomials 61 Appendix C 64 Display deviation
C.1 64
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 6 / 65
1 foreword
DKD guidelines are application documents for the requirements of DIN EN ISO/IEC 17025.
The guidelines describe technical, procedural and organizational processes that serve
as a model for the accredited calibration laboratories to define internal procedures and
regulations. DKD guidelines can become part of the quality management manuals of the
calibration laboratories. The implementation of the guidelines promotes the equal
treatment of the devices to be calibrated in the various calibration laboratories and
improves the continuity and verifiability of the work of the calibration laboratories.
The DKD guidelines should not hinder the further development of calibration methods
and processes. Deviations from guidelines and new procedures are permitted in
agreement with the accreditation body if there are technical reasons for doing so.
This guideline was created in 2005 by the Technical Committee for Force and
Acceleration in cooperation with the PTB and accredited calibration laboratories.
This revised new edition only contains an updated imprint.
It has the same content as DAkkS-DKD-R 3-9 (2010 edition). The DAkkS will withdraw
the DAkkS DKD-R 3-9 by January 1st, 2021 at the latest.
Edition: 02/2005, published by the DKD
1st new edition: 2010, by the DAkkS
2nd new edition: 2018, by the DKD, with the same content as the 1st new edition
2 Purpose and Scope

The purpose of this guideline is to define minimum requirements for the calibration
procedure and for the estimation of the measurement uncertainty in the continuous
calibration of mechanical quantity sensors. It applies to calibration objects that have to
be supplemented by an adapter provided by the calibration laboratory to form a complete
measurement chain. The adapter is part of the calibration device. This means that the
application of the guideline is fundamentally restricted to the measured variable sensor
1
without an adapter as the calibration object.
The application of the guideline presupposes that the force BNME working according
to the comparison method is able to implement both a stepped and a continuous load
profile.
Continuous calibration processes are understood here as a constant change in the
measured variable whose time behavior is, for example, ramp-shaped or sin²-shaped
(“haversine”-shaped; hav(x) = sin2 (x/2)) . However, the process is to be regarded as
quasi-static (in contrast to dynamic loading with, for example, sinusoidal or shock-
shaped time progression). It is assumed that the load time lasts from zero to the end
value of the calibration range of at least 1 minute, eg with strain gauge measurement technology.
The time-critical parameters of the continuous calibration process and the components
such as the rate of increase in load, simultaneous data acquisition, filter characteristics,
etc. influence the calibration result to a greater extent than with a step-by-step process.
1
Note: If calibration items with integrated signal adjustment are to be calibrated in the
future, this guideline can also be applied if it is ensured that the criteria described in this
guideline, e.g. with regard to synchronization and the same signal adjustment in the standard
and calibration item channels, are observed.
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 7 / 65
If the calibration facility and the metrological equipment permit a faster calibration process, the
contributions to the measurement uncertainty must be verified and quantified in each individual
case, eg in the case of piezoelectric measurement technology. The influences must be named and
specified for the purpose of comparability.
The processes and specifications in this edition of the guideline are provisionally designed to
achieve the smallest specifiable measurement uncertainty Wbmc ÿ 1·10–3 vM.
Special features specified by the calibration facility must be determined by the calibration laboratory
and taken into account when estimating the measurement uncertainty.
Note:
For the first time, the guideline describes calibration processes for which there are no
specifications in previous practice. It therefore includes metrological specifications and measures
for connection to the national standards, the transfer of the unit to the calibration laboratory and the processes t
The power BNME must be suitable in terms of start-up and control behavior.
3 Symbols and designations The symbols

and designations used in this guideline correspond to the requirements of DIN EN ISO 376 and VDI/
VDE guideline 2638. Additional symbols and designations resulting from the special nature of the
continuous process are listed in Table 1. The symbols used for the first time in the measurement
uncertainty section are listed there separately.
symbol designation
K-NME Force standard measuring device
K-BNME force reference standard measuring device
KAref Reference force transducer (part of the K-BNME)
katsd Transfer force transducer (part of the traceability measurement chain)
KG calibration item; also used as an index
NMI National Metrology Institutes
Fi power
si Measuring
index - stat signal relates to the stepwise processing of the measured
index – cont variable relates to the continuous processing of the measured variable
Index - jump refers to the sudden change in the measured variable
tB evaluation time
tf force change time
loading time in gradual expiration;
tL
Time between the start of the load and the recording of the measured value
t Load time in continuous flow
L, cont
waiting period
tw
T transducer temperature
von slew rate
fG Fundamental resonant frequency of the transducer
Sjump,i Zero-signal-corrected measured value for rapid loading and
S
unloading Zero-signal-corrected measured value for static loading and
stat,i unloading according to DIN EN ISO 376
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 8 / 65
jump,
SSÿ
ÿ
S
i = i stat, i
rel. Measured value differenceS= ÿ i=Srel,
ÿSrel,i S
stat, i stat, i
between fast and static loading unloading

dcr,E creep preload signal residual; Difference of
the output signals before loading and after
i.e0,R unloading and after waiting for the evaluation time tB ; tB must be specified
Disturbance caused by bending moments, such as eccentric force application,
lateral forces, etc.
dMo
TKC Temperature coefficient of the characteristic value

TK0 per 10 K Temperature coefficient of the zero
u signal per 10 K rel. Reversal range with fast loading and unloading
jump, i
ustat,i rel. Reversal range with static loading and unloading according to DIN EN
ISO 376 rel. Reversal margin difference
u jump,
ui i[ ÿ =
ÿ
ugh stat, ]
ÿu
i
between fast and static load rel. expanded
measurement uncertainty of the measured values of
wref the reference sensor rel. Expanded measurement
uncertainty of the measured values of the transfer
wtsd transducer Smallest measurement uncertainty that
Wbmc can be specified for static calibration Smallest measurement

wbmc,cont uncertainty that can be specified for continuous calibration Uncertainty
' interval Interpolation deviation
W
Dinp
Table 1: Symbols and designations
4 calibration item
Only those force transducers that can work with the measuring amplifier/measurement data
acquisition system of the calibration facility with regard to the electrical connection data can be
continuously calibrated as calibration objects. The calibration data are only valid if comparable
connection parameters (supply voltage, carrier frequency or DC voltage) apply in practical use
of the calibration item. Otherwise, the influencing factors must be determined by the user and
taken into account in his measurement uncertainty budget.
4.1 Calibration
capability The processing of a calibration order requires the calibration capability (suitability) of
the calibration item, ie the current condition of the calibration item should correspond to the
generally recognized rules of technology and the special requirements for a continuous
calibration process.
Before starting the calibration, quality and function tests must be carried out to ensure that the
calibration process is undisturbed.
Quality checks include, for example: Visual
inspection for damage Visual
inspections with regard to labeling/legibility of the type plate Checking
whether the documents required for calibration (technical data,
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 9 / 65
operating instructions) are available
Functional tests include, for example:

Electrical functionality
Measurement signal direction
Insulation resistance Zero
signal, e.g. B. in the non-installed state
4.2 Clamping parts, built-in parts

The built-in parts are to be regarded as part of the calibration item. The provision of the built-in
parts must be agreed between the customer and the calibration laboratory.
Tensile or compressive
force: a)Referenceforce transducer: Connection based on DIN EN ISO 376 for traceability in the
DAkkS calibration laboratory
b) Calibration object: As far as possible, provide conditions as in practical use
Coupling between a) and b) via adapters that do not falsify measured values and are low in
interference components (avoidance of bending moments, tension or deformation).
Note:
The clamping parts must be dimensioned in such a way that the calibration device is not
damaged if the clamping parts break.
4.3 Signal-carrying components in the K-BNME

The signal-carrying components are understood to mean the measuring amplifiers and the
indicators in the measuring channel for the reference sensor and the object to be calibrated.
The selection and settings of the signal-carrying components are the responsibility of the
calibration laboratory. The transmission behavior of the components must be known and taken
into account in the measurement uncertainty budget. Likewise, the same filter parameters must
be guaranteed.
Basically, a synchronization and joint triggering of the signal-carrying components for the
measurement channels of the reference sensor of the K-BNME and for the calibration object is
recommended.
Detectable systematic deviations can be taken into account by correction in the calibration
result.
A notice:
The transmission properties of adapters are e.g. Z. only determined traceable to national
standards for discrete input values, e.g. B. the voltage ratio mV/V. This is done for carrier
frequency systems as well as for DC voltage systems using so-called bridge standards.
5 Environmental conditions
The relevant environmental conditions (ambient temperature, air pressure, relative humidity)
must be monitored and documented in the calibration certificate and, if necessary, must be
taken into account in the measurement uncertainty budget. The permitted temperature range is
based on DIN EN ISO 376 from 18 °C to 28 °C. In Germany, according to the national appendix in
DIN EN ISO 376, the reference temperature is 21°C ± 2 K. During calibration, the measurement
setup must be in thermal equilibrium and the ambient temperature must be able to be kept
stable at ±1 K.
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 10 / 65
6 Performing the calibration

Basically, the calibration process is based on DIN EN ISO 376, ie the pre-loading and the
installation positions are carried out as described in the standard (see Figure 1). Under
certain conditions (see 6.1.1) these processes can be shortened.
6.1 Preparations for calibration

6.1.1 Determination of design-specific measurement uncertainty contributions
If fewer installation positions and measurement series than those described in DIN EN ISO
376 are to be carried out during the calibration process, the influencing variables required
for a measurement uncertainty assessment, such as repeatability and comparison precision
as well as hysteresis, must be determined beforehand in a sufficient number as part of a
design test calibration objects of the same type can be determined.
Hints:
- Availability of characteristic values from comparable earlier calibrations and using bau
same/similar adaptations
- Comparison of the current characteristic value with the earlier ones from the test equipment
monitoring system, for example to observe long-term instability and, if necessary, to identify damage
that has occurred in the meantime (overloads).
6.1.2 Temperature compensation

In order to be able to use the time advantages of continuous calibration, it is recommended
to let the warm-up time of the calibration objects run parallel to a running calibration by
applying an auxiliary supply voltage before installation in the K-BNME. The warm-up times
specified by the manufacturers must be observed.
6.2 calibration process

The parameters of the calibration processes (load speed, sampling rates, control settings
of the K-BNME, etc.) must be described in the calibration laboratory's QM documentation and
be available on request together with the calibration certificate.
In consultation with the customer, an informative addendum to the calibration certificate
can be created, which contains the process and the load speeds and, if necessary, waiting times.
A graphical representation is useful, e.g. according to Figure 1.
6.2.1 Preloads The 1st

preload is used for conditioning, eg for setting processes in the adaptation or for suppressing
the previous history. If further preloads are carried out, these can be used to assess the
calibration capability, for example by observing the reproducibility of the measured values.
This is particularly possible if the preload series are run with the same parameters of the
calibration process as the measurement series. The force measurement range to be calibrated
must be covered by the preload. Depending on the process, the calibration range final value
can be exceeded by approx. 3%.
6.2.2 Time management based on DIN EN ISO 376 Figure 1

shows a continuous calibration process as it can be carried out based on DIN EN ISO 376.
The time dependency of the force increase or decrease does not have to be linear, as shown,
but can also assume a sin²-shaped curve, for example. The number and the course over time
of the preloads can also be implemented differently.
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 11/65
6.2.3 Time management for shortened calibration

processes Under certain conditions (section 6.1.1) the process can be shortened to,
for example, a preload and a series of measurements with increasing and decreasing
loads or also with increasing or decreasing loads.
Figure 2a shows a sequence with a preload and an up and down series; Figure 2b
shows the simplest case, where the sequence is shortened to a preload and an up
row. It should be noted that the number of measurement uncertainty components to
be determined in a type test increases as the length of time decreases, since these can
no longer be determined directly from the measured values of a shortened process,
such as repeatability, reproducibility and hysteresis. The type tests can be carried out
according to DIN EN ISO 376, for example.
preload
Figure 1: Continuous calibration process based on DIN EN ISO 376
preload preload
f f
a) b)
Figure 2: Continuous calibration process
a) with an up and a down series b) with an upward row
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 12/65
6.2.4 Acquisition of the measured

values The acquisition of the pairs of values, consisting of the force Fref,kont and the associated
output signal SKG,kont of the calibration item, can be time-discrete or value-discrete. The time-
discrete acquisition takes place with a specified sampling frequency (typically N = 10,000 to 100,000
pairs of values). In the value-discrete recording of measured values, pairs of values (typically N = 100 to
500) are recorded for defined loads (measuring limits ).
In order to prevent aliasing, suitable filter settings must be selected.
7 Evaluation
The scope of the evaluation depends on which calibration procedure was selected (cf.
paragraph 6). When the procedure is based on DIN EN ISO 376, the measured values are processed
in such a way that all the information required for the evaluation and the assessment of the
measurement uncertainty can be obtained from the value pairs of the current calibration. In the case
of shortened calibration processes, information from previous calibrations of the calibration item
and/or type tests is required to estimate the measurement uncertainty. Deviating from the evaluation
according to DIN EN ISO 376, in which the compensatory function is only calculated from measured
values with increasing load, the measured values with decreasing load or the mean values from
both load directions can also be used depending on the application and customer requirements.
The evaluation includes the following steps (see Figure 3):
1. Determine the output signal SKG of the calibration item at the required interpolation points Fref,i
either: a)
determination by linear interpolation of adjacent pairs of measured values (SKG,kont, Fref,kont)
or b) determination using best fit functions (higher degree) over all
Measured value pairs, separately for measured value pairs in the direction of increasing and decreasing
Load for each mounting position (see appendix for more information on how to determine these
balancing functions if random deviations need to be reduced
b)
2. Provision of measurement uncertainty contributions (e.g. repeatability, reproducibility, hysteresis)

from design tests. When carrying out several series of measurements: Determination of mean
values and measurement uncertainty contributions in analogy to DIN EN ISO 376, insofar as
these can be determined from the pairs of measured values.
3. Determination of the transfer coefficient from the pairs of values at the interpolation points of the
upward and/or downward series, which were calculated under point 1, preferably as the slope of
a first-order fit function through zero using the (Gaussian) method of least squares (other
methods see Section 8.5). Based on DIN EN ISO 376, compensation of a higher degree is also
possible.
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 13 / 65
When using a 1st degree compensation:
4. Set up a value table with the pairs of values at the reference points (which were
calculated under point 1), the associated expanded measurement uncertainties W,
the interpolation deviations Dinp and the uncertainty intervals W'. The uncertainty
intervals for the transfer coefficient (slope of the first-order regression function, one-
value specification) result from the addition of expanded measurement uncertainty
and interpolation error at the reference points.
5. Representation of the transmission coefficient (single-value specification) with

uncertainty intervals and specification limit(s) in the calibration range. One or more
self-determined specification limits can be assigned to the transmission coefficient
(the single value specification). The force range for which the specification limits
are valid must be specified.
6. Specification of the complete measurement result consisting of transmission

coefficient (one value specification) and specification limit(s) (this is larger than the
uncertainty interval; see Section 8.5)
The output signal of the calibration item depending on the force acting on it can
alternatively be described by a table of pairs of values, eg in analogy to DIN EN ISO
376. The pairs of values make it possible to use better fitting functions instead of
specifying a single value.
In order to illustrate the effect of the polynomial formation after step 1b, Appendix B
shows a comparison of the results obtained when a transmission coefficient (one-value
specification) is calculated directly from the recorded value pairs (Fref,kont; SKG,kont)
and after reduction of random deviations is determined by using a polynomial formed
according to Annex B.
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 14 / 65
Installation position 240°

Installation position 120°
data Installation position 0°
collection cont
KG,
FS;
ref,cont KG, cont cont
KG,
S
S
cont
KG, cont
KG,
S f
S Ref, cont
f Ref, cont
F ref,cont F ref,cont
1) Determination of pairs of values at interpolation points by: b)

a) Linear interpolation between two adjacent Calculation using a higher degree
measured values polynomial with all measured values
Free, cont Polynomial, ascending,
support points dividing x° Polynomial,

descending, dividing x° grid points
KG KG
S
S
F ref,i F ref,i
,
2) Determination of mean values and measurement uncertainty contributions at the reference points
as far as possible from the series of measured values; otherwise provision from type tests.
3) From the pairs of values at the interpolation points, determination of a

first degree compensation of a higher degree
KG KG
S
S
F ref,i F ref,i
4) Set up the table of values

´ ´
f ref, i S KG W D
inp W f ref, i S KG W D
inp W
with W
from model with W
out model
“Transmission Coefficient” "Power"
W'
5) Uncertainty intervals at the 5) Uncertainty intervals at W'
support points + specification limits the reference points
´
W'
W
F ref,i F ref,i
6) full Measurement result: Transmission 6) Table of values +

coefficient. or single value plus specification limit Coefficients of a Polynomial
Figure 3: Process of an evaluation

DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 15 / 65
8th measurement uncertainty

8.1 influencing factors
When setting up the evaluation model, it is recommended to list the influencing factors and to
present them graphically. The cause-effect diagram is suitable for this; it visualizes the relationship
between the cause (excitation variable) and the effect (result variable, measurement result) under
the influence of disturbance variables.
Note:
The smallest measurement uncertainty that can be specified is determined beforehand using
transfer force transducers (see
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 16 / 65
Annex A).
smallest measurement uncertainty that can Temperature change with continuous load in continuous
be specified with continuous loading operation
power
calibration certificate calibration certificate resolution
interpolation deviation
temperature temperature reversal margin
Comparison precision (rotation)
long-term instability long-term instability repeatability
zero drift
indicator customizer calibration item
Figure 4: Influencing variables when determining the measurement uncertainties associated with the
measured values of the calibration item
Determining the measurement uncertainty does not replace the qualifications and experience of the
measurement technician, e.g. B. at the
Selection of the appropriate measurement method

Design of the measurement setup
Carrying out the measurements
Evaluation of the measurements
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 17 / 65
Other sources of measurement errors are described in Table 2 below:
source cause
Measuring device Calibration interval expired,

internal and external adjustments,
Resolution, noise, drift, shielding technology
Environmental conditions Ambient temperature, air pressure, relative humidity,
Vibration, shock, surrounding medium
Circuit/experimental setup impedances, lines,
Thermal voltages, supply voltages, mechanical
adaptations, installation locations, positions,
Masses, gravitational acceleration
observer Writing, reading errors, presence of the
Observer's
programming incorrect file accesses,
incorrect constants, incorrect parameter transfer
Table 2: Sources of measurement deviations and measurement uncertainties [19]
Basically, it must first be ensured that all systematic deviations are recognized and that the
corrections are made for them. The remaining unknown measurement deviations are to be taken
into account in the measurement uncertainty budget in the form of estimated measurement
uncertainty contributions.
Notes: 1.
During continuous operation, the KAref can heat up due to the constantly repeating elastic
deformation of the range spring , which can lead to changes in the characteristic values. In order
to assess this influence, comparative measurements can be carried out with a transfer force
transducer before and after a sufficiently large number of load cycles (calibration runs).
2. If the temperature of the KAref in calibration mode changes by more than ±1 K compared to
the reference temperature, the measurement result can be corrected. The consideration of the
temperature influence in the measurement uncertainty budget is then reduced to the uncertainty
when determining the sensor temperature.
8.2 model of the evaluation

The models of the evaluation of force calibrations are formulated as linear product models using
relative measurement uncertainties.
In general, the following applies:
N
Sizes: Y q X pÿ
=ÿ p
; := ± 1 (8.1)
ii i
i=1
N
estimated values: =ÿ
yq x ÿ pi
i
; pi 1 = ± (8.2)
i
8.2.1 "Transmission coefficient" model In the

detailed representation - equation for determining the transmission coefficient of a force
transducer for individual force values with "best" estimated values - the expression becomes:
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 18 / 65
V no
(GM ÿ
) 5
E KG =
Sp
ÿ
ÿ K i
F
bmc, cont i=1 (8.3)
5
With ÿ KKKKKK
i
=ÿ
T crush
ÿ
representative
ÿ
red
ÿ
rev
i=1
2
ECG result size; Transmission coefficient (sensitivity) of the KG
34
VNo Output voltage indicator
G Gain coefficient of the KG amplifier
vs Supply voltage for the KG
Fbmc,kont Force value set by the K-BNME in continuous operation
correction factor
KT
rel. Temperature change with continuous load in continuous operation
concentration camp
Correction factor rel. zero point deviation correction factor rel.
crepe Repeatability Correction factor rel. Comparison precision Correction
bread factor rel. reversal margin
crev
Note: The
measurement uncertainty contributions of the influencing variables emitter, amplifier and supply voltage can
generally be taken from the calibration certificates of the devices. In the guideline VDI/VDE/DGQ/DKD 2622
Calibration of measuring equipment for electrical variables in sheet 3 digital multimeter and sheet 20 DC voltage
measuring amplifier there are instructions on this.
Correction Factors:
ÿ ÿx ÿ
K i = + 1ÿ ÿ i With
xi
ÿÿ
ÿx i unknown measurement deviations and

E[ÿx =] i0 expected value
8.2.1.1 Measurement uncertainty

contributions Due to the selected, linear product model and the application of relative
5 ÿ
ci 1 = ±
measurement uncertainties, the sensitivity coefficients are always ).
equal to 1 (
2
Result variable
3 N' input variables for determining the result variable 4 N input N'<N
variables for determining the measurement uncertainty c is the N'<N
*
5 The sensitivity coefficient factor
i with which the relative uncertainty w x( i) is converted into the relative
uncertainty contribution wy i ( ) . It is related to the sensitivity coefficient c (see i
ÿ xi
Eq. 4.3 in DAkkS-DKD-3) as follows: ci =ÿ
ci
y
DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 19 / 65
The contributions wi(y) to the uncertainty of the result variable thus correspond to the uncertainties of the
input variables w(xi):
w (E KG ) =
2 2 2
) 2
w G( w) V+w F +( +Spbmc,cont
2 2
( T ) + ( +) + 2
( 2 2 (8.4)
w V( no ) ( ) wKwKwKwKwK crush representative )+ ( red )+ ( rev )
The associated distribution functions and rel. Variances can be found in Table 3 below:
probability distribution
random variable rel. Variance w²(xi)
2
VNo normal VNo
()w
2
G normal w G( )
2
vs normal Vsp()w
2
fbmc,cont normal Fbmc,cont
()w
2
2 a
KT rectangle T w( K T=)
3
2 2 acrush
concentration camp
rectangle wK(= crush
)
3
2
crepe rectangle ( =)
awK
2 representative
3
representative
2 2a
bread U-shaped wK(= red
) red
2 2 arev
crev rectangle wK(= rev )
3
Table 3: Distribution functions and variances of the influencing variables

DKD-R 3-9
Output:
Revision: 0
10.7795/10.7795/550.20180823J
Side: 20 / 65
8.2.1.2 Uncertainty Analysis The

knowledge about the input variables is preferably summarized in a table (Table 4).
half-width of uncertainty
ongoing Estimate insecurity
size the divider variance
No.
distribution contribution
sensitivity coefficient
Xi xi a w(xi) ci * wi(y) 2 w (y)
1 VNo x1 v 2 w V( Anz ) w 1 E ( KG
wVanz E(
w2no KG )
) V
2 G
x 2/ 2 G( ) w -1 wGE( KG ) ) wG2EKG )
3 vs x3 v 2 V( Sp ) -1 wVSp
E KG wV2E
KG )
( Sp
((
4 fbmc,cont x4kN _ 2 w F( bmc,kont) 3 -1 ( E KG

w Fbmc,kont E cont (
wF2bmc, KG )
)
5 KT 1 aT w K( T ) ( ) w Kzer 1wE T ( CG ) wT2E
KG )
6 concentration camp 1 a crush 3 ( ) w( ) w

Krep 1wE Zer CG ( ) wZer
2E
KG )
7 crepe 1 arep
3 w Krot
Krev( ) 1wE rep CG ( )
w 2E (rep
KG
(( )
8th bread 1 a red
2 1wE red CG ( ) wred
2E
KG )
2
( rev
(w KG E )
9 a
crev rev 1f rev
1 3 E CG ( )
N
mV/V rel. standard measurement uncertainty
ECG y w (E KG =ÿ) wi2E( KG
kN w i= 1 )
mV/V rel. extended measurement

ECG y WE( KG ) =kwÿ E( KG )
kN uncertainty W
mV/V
Indication of the complete measurement result E KG
=y WE( ±) KG
kN
Table 4: Uncertainty analysis for the transmission coefficient model
However, the measurement uncertainty must be estimated for each calibration value, ie for each specified
exposure level. The following measurement uncertainty budget (Table 5) serves to provide a clear overview:
DKD-R 3-9
Continuous calibration of force transducers using the comparison Output: 1
method https://doi.org/10.7795/550.20180823J Revision: 0
Side: 21/65
8.2.1.3 Measurement Uncertainty Budget

Rel. standard measurement uncertainty wi(y)
(measurement uncertainty contributions associated with the result due to influencing variables)
temp. Zero point reversal
power transmission no adapt
adjust (cont. exposure, bmc
deviate from continuous exposure drive) f0 # span span
coefficient span
v
V G vs fbmc,cont b' b
kN (mV/V)/kN
for fmin
...
for Fmax
Rel. standard measurement uncertainty wi(y) Rel. standard Rel. expanded

(combined measurement uncertainty contributions associated measurement measurement
power transmission coefficient with the result due to influencing variables) uncertainty w(y) uncertainty W(y) (k=2)
cient smallest specifiable object to
measurement uncertainty procedure In total
be calibrated
fbmc,cont
kN (mV/V)/kN #
for fmin
...
for Fmax
Table 5: Measurement uncertainty budget for the transfer coefficient model (all reference points in the range)
Continuous calibration of force DKD-R 3-9

transducers using the Output: 09/2018
comparison method https:// Revision: 0
doi.org/10.7795/550.20180823J Side: 22 / 65
8.2.2 "Force" model (polynomial coefficients of a 3rd degree compensation, DIN EN ISO 376)
In the detailed representation of the linear product model - conditional equation with the
"best" estimated values - this becomes the expression:
8th
=
F F bmc,
KG
cont
ÿ
ÿ K
i
i=1
8th
(8.5)
With ÿ KKKKKKKKK
i
=ÿ
T adjust
ÿ
no
ÿ
crush
ÿ
representative
ÿ
red
ÿ
inp
ÿ
rev
i=1
6
FKG result size; force acting on a KG
78
Fbmc,kont force value set by the K-BNME during continuous operation
correction factor
KT
rel. Temperature change with continuous load in continuous operation
canp Correction factor adaptor
can Correction factor indicator
concentration camp
Correction factor rel. zero drift
crepe correction factor rel. repeatability
bread correction factor rel. comparison precision
Kinp correction factor rel. interpolation deviation

crev correction factor rel. reversal margin
Correction Factors:
ÿ ÿx ÿ
i
Ki = + ÿ ÿ1 With
x
i ÿÿ
ÿx i unknown measurement deviations and

E[ÿx =] i0 expected value
6
result size
7 N' input variables for determining the result variable N'<N
8 N input variables for determining the measurement uncertainty N'<N

doi.org/10.7795/550.20180823J Side: 23 / 65
8.2.2.1 Contributions to measurement

uncertainty Due to the selected linear product model and the consideration * of relative measurement
i = ±1 ).
uncertainties, the sensitivity coefficients are always equal to 1 ( c The contributions wi(y) to the uncertainty
of the result variable therefore correspond to the uncertainties of the input variables w(xi):
w (F KG ) =
2 2
w K(w K
2 2 2
(
2 2 2 2 (8.6) )
( wK + w K (w Know)K w K w
( Kcrush
++ ) +inp )+ ( )+ )+ ( )+
w F bmc,
cont ) )
adjust representative
red ( rev ( T
The associated distribution functions and rel. Variances can be found in Table 6 below:
2
fbmc,cont normal ( ) w Fbmc,cont
2
2 a
KT rectangle T w( K T=)
3
2
canp normal chap( () )ww
2
can
normal chance
2 2 acrush
concentration camp rectangle wK(= crush
)
3
2a
2 rep
crepe rectangle wK (= )
3
representative
2 2 ared
bread U-shaped ) w(K =
red
2
2
2 a
Kinp triangle inp (w K =)inp
6
2 2 arev
crev rectangle
K =( ) w
rev
3
Table 6: Distribution functions and variances of the influencing variables


doi.org/10.7795/550.20180823J Side: 24/65
8.2.2.2 Uncertainty Analysis
Half
uncertainty
ongoing Estimate width insecurity
size divider variance
No. of the
contribution
distribution *
Xi xi a w(xi) c i wi(y) 2 w (y)
2
1 fbmc,cont x 2 1 wbmc,kontfKG ( ) w KG (F
bmc, kont )
w
2
2 KT 1 aT 3 1 wTE( KG ) wT E )
KG
2
3 canp 1 F( bmc,kont) w 1wF anp KG ( ) w F anp
KG( ( )
2
4 can 1 2
K( T ) 2 w 1 wanz
F KG ( ) wnum
F KG )
2
5 1 a 3 1 ( ( wzerKG )
concentration camp crush
w K( anp
anz ) wzer
FKG ( ) wrep
2
6 1 a 3 1 F
( wrep )
crepe rep ()w
Kzer ()w FKG (( ))
wrot FKG KG
2
7 1 a 2 1 F
( wrot )
bread red w Krep
Krot (( )) FKG winp
( ) wrev KG
2
1 a 6 1 FKG ( winpF FKG )
8th
Kinp inp ( )wKinp
wKrev
2
9 crev 1 arev 3 1 ( wrev F
KG )
N
rel. standard measurement uncertainty
FKG y w (F KG )=ÿ w i2
w
i= 1
rel. extended measurement

FKG y WF( KG
kw) F= ÿ ( KG )
uncertainty W
Indication of the complete measurement result FKG = ±y W F( KG )
Table 7: Uncertainty analysis for the force model (polynomial

coefficients of a 3rd order fit function)
DKD-R 3-9
Side: 25/65
8.2.2.3 Measurement Uncertainty Budget

power temp. Interpolation reverse
zero point (cont. exposure,
span ons deviation
spancontinuous
(red) operation)
bmc, cont ment customizer indicator span
kN #
at least
Max.
rel. Standard measurement uncertainty wi(y) rel. Standard measurement rel. Expanded measurement
(combined measurement uncertainty contributions uncertainty w(y) uncertainty W(y)
power assigned to the result due to influencing variables) smallest (k=2)

specifiable measurement uncertainty (bmc, kont) object to
procedure In total
be calibrated
kN #
at least
Max.
Table 8: Measurement uncertainty budget for the polynomial coefficients model


comparison method Revision: 0
https://doi.org/10.7795/550.20180823J Side: 26 / 65
8.3 Visualization of the measurement uncertainty

budget If you visualize the uncertainty contributions (variances) in a bar chart, with the
contributions sorted in descending order, you can see in an impressive way which influencing
factors require action if the measurement uncertainty is to be reduced.
Contributions of the smallest power level
80%
70%
60%
50%
Portion
40%
30%
20%
10%
0%
bmc
(red) (rep)
different
span span
span
reversal indicator point
zero customizer
influencing factor
Figure 5: Representation of the variance shares (square of the uncertainty contributions) in the
Total variance (square of the standard measurement uncertainty) in
descending order.
8.4 uncertainty interval
The relative uncertainty interval W' is made up of the amount of the systematic measurement
deviation and the relative expanded measurement uncertainty (k = 2). Due to the systematic
part, the rectangular distribution is assigned to the uncertainty interval as the form of
probability distribution. In the case of the transfer coefficient model, the interpolation
deviation from the fit function of the 1st degree is such a systematic measurement deviation
(see single-value specification):
' ÿ S
inp
W = +W (8.7)
S
8.5 One-value specification as a calibration

result A measurement uncertainty is always associated with a measured value. In the case of
calibrations with different and usually equidistantly distributed values within a range, tables
of results are thus obtained (e.g. Table 5.). The user of the calibration item, on the other hand,
often only works with one value for the entire validity range of the calibration, the so-called
single-value specification. This one-value specification is, for example, the transmission coefficient
DKD-R 3-9
Continuous calibration of force
doi.org/10.7795/550.20180823J Side: 27 / 65
(Sensitivity) of a measurement sensor for the measuring range. According to the rules of the
GUM, however, no measurement uncertainty can be specified for the one-value specification; it
is replaced by a statement of conformity.
For this, the determination of specification limits is taken into account

the systematic measurement deviation
(of the individual value from the single-value
specification) and the associated expanded measurement uncertainty
(of the individual value).
The upper and lower specification limits are approximated to the largest uncertainty interval
(sum of systematic measurement error and expanded measurement uncertainty); preferably
with a small safety margin, so that smooth numerical values result.
The self-determined specification limits determined in this way must be taken into account as
a measurement uncertainty contribution in the measurement uncertainty budget of the
application. Due to the systematic part of the uncertainty interval, the contribution is assigned a rectangular dis
senior
In the case of calibration objects with a nominal characteristic value adjusted by the
manufacturer (e.g. with resistive sensors), the specification limits can alternatively also be
determined from the associated characteristic value tolerance.
Various methods are currently in use to describe the transmission behavior of the calibrated
transducer, characterized by a number of determined transmission coefficients, using a single
value: 1. The calculation and use of the mean value of the determined transmission coefficients
as a one-value specification.
2. The determination of a regression line from the pairs of measured values determined (display
of the calibration item and the set value of the measured variable) using the slope of the
line as a one-value specification. The regression calculation can be subject to various
criteria: • The regression line is selected in such a way that the start and end point of the
measured characteristic curve coincides with the regression line (referred to in the literature
as the fixed point method or limit point setting). • The regression line is selected in
such a way that the starting point of the measured characteristic coincides with the
starting point of the regression line (referred to in the literature as the minimum method or
starting point setting). The gradient of the straight line is then chosen in such a way
that the deviations from the measured characteristic curve fulfill a specific minimum
principle. As a minimum principle, for example, the condition can be used that the
maximum deviation is minimal (Chebyshev approximation; also known as the “best
straight line”, cf. ANSI/ISA S37.1 [20]) or the sum of the squares of the deviation is
minimal (according to Gauss)
• The regression line is selected in such a way that the deviations from the measured
characteristic fulfill a minimum principle. (Also referred to as the tolerance band
method in the literature.)
Continuous calibration of DKD-R 3-9

force transducers using the Output: 09/2018
fixed point method minimum method tolerance band method
measured
curve measured
measured curve
f(x) f(x) curve f(x)
compensation
compensation compensation
straight
straight straight
x x x
Figure 6: Various methods for obtaining a one-value statement
deviation deviation deviation
x x x
Figure 7: Comparison of the resulting deviations between the measured

Characteristic curve and the regression line with reference to the final
value of the regression line
deviation deviation deviation
x x x
Figure 8: Comparison of the resulting deviations between the measured

Characteristic curve and the regression line when referring to the (respective) value of the
regression line
The different methods lead to different results.

It is therefore recommended that the one-value specification be given as the slope of a
straight line through zero, whose coefficients are calculated by linear adjustment using
the (Gaussian) method of least sum of squares (minimum method).
In order to ensure comparability of the calibration results, the method used to determine
the initial value must be specified in the calibration certificate.
8.6 Visualization of the calibration result In

addition to the tabular representation, the calibration results can be visualized further
for purposes of illustration.

8.6.1 Deviation and measurement uncertainty

The specification of measured values is supplemented by a deviation diagram (here:
interpolation deviation) with data related to the measured value (of average). If
specification limits (here: class limits) are known, the diagram allows a decision to be
made at a glance as to whether the calibration item complies with the specification limits.
Shown are:
Calibration
range rel. interpolation deviations
v. M.rel. Expanded measurement uncertainty (k = 2) at
each reference point Specification limits Limits of
validity
In addition, the “maximum relative deviation from M." and for purposes of comparison
with data sheet information, which usually refers to the final value, also the "maximum
relative deviation based on the final value (of E.)" (often referred to as linearity, cf. ANSI/
ISA S37.1 [20]) indicated.
Note:
The information is necessary to enable the user of the calibration item to calculate his
measurement uncertainty during use. The content and structure of the deviation diagram is
independent of the calibration of an indicating calibration item or a calibration item with an
unnamed scale.
0.30%
0.25%
0.20%
0.15%
0.10%
Measurement
uncertainty
average)
k=2)
((of
0.05%
0.00%
interpolation
deviation
with
ext.
Rel.
rel.
-0.05%
-0.10%
-0.15%
-0.20%
-0.25%
-0.30%
0 2 4 6 8th 10 12
Force in kN
Interpol. dev. measurement uncertainty class specification limits validity limits
Figure 9: Deviation diagram with specification and validity limits related to the
measured value
Note: The
lower limit of the validity range has been arbitrarily set to 10% of the calibration range in this
example.

8.6.2 Single value The

transmission coefficients and the associated expanded measurement uncertainties are
shown in comparison to the specification limits according to the manufacturer's
information or to the self-determined specification limits.
0.2025
0.2020
0.2015
0.2010
measurement
Transmission
uncertainty
coefficient
expanded
(mV/
with
(=2)
kN
V)/
in k
0.2005
0.2000
0.1995
0.1990
0 1 2 3 4 5 6 7 8th 9 10 11
Force in kN
transmission coefficient Extended measurement

Single-value indication B, transmission coefficient for range uncertainty self-determined specification limit 0.6% f.s. B
validity limits
Figure 10: Transmission coefficients with ext. Measurement uncertainty, one-value

specification (transfer coefficient for the range) and self-determined
specification limit.
0.0015
0.0010
0.0005
Uncertainty
interval
(mV/
kN
V)/
in
0.0000
-0.0005
-0.0010
-0.0015
0 1 2 3 4 5 6 7 8th 9 10 11
Force in kN
uncertainty interval Self-determined specification limit 0.6% f.s. B validity limits
Figure 11: Uncertainty interval and self-determined specification limit

DKD-R 3-9
Continuous calibration of
https://doi.org/10.7795/550.20180823J Side: 31/65
0.2050
0.2040
0.2030
0.2020
0.2010
measurement
Transmission
uncertainty
coefficient
expanded
(mV/
with
(=2)
kN
V)/
in k
0.2000
0.1990
0.1980
0.1970
0.1960
0 1 2 3 4 5 6 7 8th 9 10 11
Force in kN
transmission coefficient Extended measurement uncertainty

Single-value indication B, transmission coefficient for range Self-determined specification limit 0.2% f.s. B * E./M.
validity limits
Figure 12: like Figure 10, but with hyperbolic, self-determined specifications
border
0.0040
0.0030
0.0020
0.0010
Uncertainty
interval
(mV/
kN
V)/
in
0.0000
-0.0010
-0.0020
-0.0030
-0.0040
0 1 2 3 4 5 6 7 8th 9 10 11
Force in kN
uncertainty interval Self-determined specification limit 0.2% f.s. B*E./M. validity limits
Figure 13: like Figure 11, but with hyperbolic, self-determined specifications
border

doi.org/10.7795/550.20180823J Side: 32 / 65
9 Information in the calibration certificate

Basically, DAkkS-DKD-5 “Instructions for creating a DAkkS calibration certificate” must be observed.
In addition, information is required that characterizes the continuous process.
Clear reference to the continuous calibration process according to DKD-R 3-9 with eg: -
rate of increase or increase duration of the force per load - description of the data
acquisition - number of pairs of values that are used for the evaluation
-
if the calibration procedure does not correspond to DIN EN ISO 376, information on
previous loads, number of measurement series, installation positions, etc. must be
provided, if necessary with visualization according to Figure 1 or Figure 2; especially for
processes according to customer requirements
Information on the adapter/measuring amplifier:

- Transmission coefficients/gains
- Supply voltage (with/without voltage feedback), carrier frequency or
DC voltage;
- Filter and filter characteristics
- Note that direct use is only possible when using adapters of the same design
of the calibration certificate information is possible.
Cable:
- type, eg highly insulating, type of shielding - for
transducers in a bridge circuit, eg 4- or 6-wire circuit; with piezoelectric
sensors, eg cable capacity - length
built-in parts
Environmental conditions
Zero signal without force introduction parts at nominal supply voltage
Type-specific contributions to the measurement uncertainty if the calibration processes are

shortened compared to DIN EN ISO 376.
The calibration result must be specified

- preferably the transmission coefficient as the slope of a regression line
through the coordinate origin (single-value specification)
Note 1: The origin of the measured values on which the adjustment calculation is based
must be specified (e.g. measured values in the direction of increasing load or mean
values from measured values in increasing/decreasing load)
Note 2: The calculation method must be stated, see Section 8.5 -
self-determined specification limit(s) and associated validity limits Note 1: The user of the
calibration item can take into account the self-determined specification limit as an input
variable in his measurement uncertainty budget, assuming a rectangular distribution
Note 2: A visualization of the results is recommended as described in paragraph 8.6.

10 Records in the calibration laboratory Records on the

calibration setup with assignment to the order are to be kept and kept in the calibration
laboratory, e.g.:
Additional information relevant for the traceability, e.g. K-BNME and

reference force transducer used Temperature of the reference force
transducer during calibration Software version for computer-aided
operation Description of the installation position for KAref and KG,
e.g. through the direction of the cable outlet Information about the operator Slots
of the measuring amplifiers in multi-channel systems
11 conformity
If the uncertainty interval or the transmission coefficients with associated expanded
measurement uncertainty are within the required specification limits, then conformity
according to DAkkS-DKD-5 can be confirmed. The area of validity (force area) must be
specified.
When assessing compliance with the required specification limits, their origin must be
specified, e.g. class limits according to DIN EN ISO 376, manufacturer-specific (internal)
Information according to the data sheet. If it is about self-determined specification limits
from the current data, this must be indicated.
12 calibration mark
The calibration item receives a calibration mark. In the case of calibration items that
consist of several separable components, each component is given an identical mark.
If the item to be calibrated is stored in a container for protection, it must also be
assigned a brand of the same name to make it easier to find.
If the dimensions of the calibration item do not allow a calibration mark to be attached,
a corresponding note must be included in the calibration certificate.

13 Scripture
13.1 Technical rules for
force DIN EN ISO
1. 376: Metallic materials - Calibration of force measuring devices for testing
testing machines with uniaxial stress, Beuth Verlag Berlin, 2003 edition VDI /
VDE 2638: Parameters for force transducers, terms and definitions, April 1989
2.
3. DKD-R 3-3: Calibration of force gauges, PTB, Braunschweig, 2nd new edition
2018
Torque DIN
4. 51309: Calibration of torque measuring devices for static torques, 1998
EA-10/14: EA Guidelines on the Calibration of Static Torque Measuring
5. Devices European co-operation for Accreditation, rev.00, June 2000, pp 20;
superseded by EURAMET cg-14 Version 2.0 (03/2011), https://www.euramet.org/
publications-media-centre/calibration-guidelines/
6. DKD-R 3-5: Calibration of torque measuring devices for static alternating

torques, PTB, Braunschweig, 3rd new edition 2018 DKD-R 3-7: Static calibration
7. of indicating torque wrenches, PTB, Braunschweig, 2nd new edition 2018 DKD-
R 3-8 : Static calibration of calibration devices for torque wrenches, PTB,
8th. Braunschweig, 2nd new edition 2018 Measurement uncertainty, general: Guide
to the Expression of Uncertainty in Measurement (GUM)
9.
First Edition 1993, ISO, Geneva CH, ISBN 92-67-10188-9
10 DIN V ENV 13005: Guidelines for specifying the uncertainty when measuring,
Beuth Verlag Berlin, 1995, ISBN 3-410-13405-0 (German translation of the "
Guide to the Expression of Uncertainty in Measurement"); EA-4/02 (formerly
EAL-R2): Expression of the Uncertainty of Measurement in Calibration - including
supplement 1 and 2, European co-operation for Accreditation, December 1999,
79 p., http://www.european accreditation.org
11. EURACHEM / CITAC Guide: Quantifying Uncertainty in Analytical Measurement
Second Edition, 2000, pp. 120, http://www.european-accreditation.org DAkkS-
12. DKD-3: Specification of the measurement uncertainty for calibrations
(German version of the publication EA- 4/02: "Expression of the Uncertainty of
Measurement in Calibration "), German accreditation body DAkkS, Braunschweig,
new edition 2010 DAkkS-DKD-3-E1: Specification of the measurement uncertainty
13. in calibrations, supplement 1, examples (German version of the publication EA-
4/02-S1: Expression of the Uncertainty of Measurement in Calibration,
Supplement 1 -Examples), German Accreditation Body DAkkS, Braunschweig,
new edition 2010 DAkkS-DKD-3-E2: Specification of the measurement uncertainty
in calibrations, Supplement 2 - Additional Examples- (German version of
14 publication EA-4/02-S2: Expression

doi.org/10.7795/550.20180823J Side: 35 / 65
of the Uncertainty of Measurement in Calibration, Supplement 2 -Examples), German

accreditation body DAkkS, Braunschweig, new edition 2010
15 DIN 1319-3: Fundamentals of metrology Part 3: Evaluation of measurements of a single
measurand, measurement uncertainty, Beuth Verlag Berlin, 1996
16 DIN 1319-4: Basics of measurement technology Part 4: Evaluation of measurements,
Measurement uncertainty, Beuth Verlag Berlin, 1999
17 VDI/VDE 2620: Uncertain measurements and their effect on the measurement result
Beuth Verlag Berlin, draft December 1998
Measured variable Geometrical variables

18 VDI/VDE/DGQ/DKD 2618: Test equipment monitoring Instructions for monitoring measuring
equipment for geometrical variables, Part 1.2: Measurement uncertainty, Beuth Verlag
Berlin, December 2003
Measurand Electrical quantities VDI/

19 VDE/DGQ/DKD 2622: Calibration of measuring equipment for electrical quantities, Sheet 2:
Methods for determining measurement uncertainty, Beuth Verlag Berlin, May 2003 ANSI/ISA-
20 S37.1-1975 (R1982): Electrical Transducer Nomenclature and Terminology Instrument
Society of America (ISA), Research Triangle Park, North Carolina
Measured variable force and torque:

21 EA-10/04 (formerly EAL-G22): Uncertainty of Calibration Results in Force Measurements,
European co-operation for Accreditation, August 1996, pp 16, replaced by EURAMET
cg-4 Version 2.0 (03/ 2011), https://www.euramet.org/publications-media-centre/
calibration-guidelines/
13.2 literature
Transducer properties, especially creep:
Kraft
22 Paetow, J.: The effect of time response of force transducers on their transfer proper ties in
force comparisons, IMEKO, Amsterdam 1986 Bethe, K.: After-effects in load cells, IMEKO,
23 Turin 1994 Peters, M.: Known and unknown influences on the uncertainty in force
24 measurement in Basic Metrology and Applications, Libreria Editrice Universitaria Levretto &
Bella, Torino 1994, pp. 88-94 Kumme, R.: Investigation of a direct method for the dynamic
calibration of force measuring devices - a contribution to reducing the measurement
25 uncertainty (Dissertation), excerpt from PTB report MA-48, Braunschweig, August 1996
Bartel, Th. W.; Yaniv, SL: Creep and Creep Recovery Response of Load Cells Tested
According to US and International Evaluation Procedures, J. Res. Natl Inst.
26
hours Techn. 102, 349 (1997)

27 Saffar, J.; Camarano, D.; Ferreira Marques, E.: Behavior of elastic force-measurement
instruments under different loading regimes, IMEKO, Tampere 1997
Torque Röske,
28 D.; Peschel, D.: Investigations into the alternating torque calibration of torque transducers,
IMEKO, Tampere 1997

29 Peschel, D.; Brüge, A.: Calibration of torque measuring devices – step-by-step or

continuous procedure, IMEKO, Korea 1998 Peschel, D.; Röske, D.: Determination
30 of creep value using short-term creep, IMEKO, Osaka 1999
Properties of piezoelectric transducers: Mack,

31 O.; Kumme, R.: Quasi-static investigations of piezoelectric force measuring devices,
PTB Braunschweig, annual reports 1998, pp. 171-172 Kumme, R.; Mack, O.; Bill, B.;
32 Haab, HR; Gossweiler, C.: Investigation of Piezoelectric Force Measuring Devices
in Force Calibration and Force Standard Machines, IMEKO TC3, September 2001,
Istanbul, pp. 92-103 Mack, O.: New Procedures to Characterize Drift and Non-Linear
33 Effects of Piezoe Electric Force Sensors, IMEKO TC3, September 2001, Istanbul, pp.
141-148 Kumme, R.; Mack, O.; Bill, B.; Gossweiler, Ch.; Haab, HR: Dynamic
34 Properties and Investigations of Piezoelectric Force Measuring Devices, IMEKO
Celle, 2002, VDI Reports No. 1685, pp. 161-171 Mack, O.: Investigations of the
Influence of Disturbing Components on a Piezoelectric Force Transducer, IMEKO
35 Celle, 2002, VDI Reports No. 1685, pp. 417-424 Gossweiler, C.: A proposal for
qualifying the continuous calibration of force sensors by comparison with the
36 gradual calibration according to EN 10002-3, Kistler Instruments AG, February 2003,
17 p.
Measurement uncertainty,
37 general: Weise, K. and Wöger, W.: A Bayesian theory of measurement
uncertainty, report PTB-N-11 PTB, Braunschweig and Berlin, August 1992
38 Weise, K. and Wöger, W.: Measurement uncertainty and measurement data
evaluation, VCH Weinheim, 1999, ISBN 3-527-29610-7 Wöger, W.; Kessel, W.;
39 Sommer, K.-D.; Siebert, BRL; Kind, D.: Focus on measurement uncertainty, reprint
from issue 3 and issue 4 of PTB-Mitteilungen 111 (2001)
40 Adunka, F.: Measurement uncertainties - theory and practice, Vulkan-Verlag Essen,

2000, ISBN 3-8027-2194-2
Measurement uncertainty, force and torque:

41 Sawla, A.: Uncertainty of measurement in the verification and calibration of
the force-measuring systems of testing machines, Proceedings of the Asia-Pacific
Sym posium on Measurement of Force, Mass and Torque (APMF 2000), Tsukuba,
Japan, November 2000, P. 7 - 14 Sawla, A.: Uncertainty Scope of the Force
42 Calibration Machines, Proceedings (Vol ume III, P. 253-258) of the XVI IMEKO World
Congress (IMEKO 2000) Vienna, From tria, September, 2000 Peschel, D.:
Determination of the smallest specifiable measurement uncertainty of torques
43 using a torque reference standard, draft, working paper FL 1.13, May 2000

Appendix A
A.1 Traceability of the force BNME A.1.1

General The traceability of the force reference
standard measuring device (K-BNME) is based on the force standard measuring devices
(K-NME) of the National Metrology Institutes (NMI) (see Figure 14 and .15). Since it is
e.g. At the moment there is no national standard for calibrating force transducers with
continuously increasing or decreasing load over time, the traceability is carried out
using transfer force transducers KAtsd as with static loads.
In addition to these step-by-step calibration processes, the suitability of reference and

transfer force transducers must be verified by further measurements (see Figure 15 and
Section A.3). This includes determining the unloading creep and the influence of the
loading speed on the measured values or the characteristic value and, if applicable, the
reversal range.
As with the calibration according to DIN EN ISO 376 or comparable processes, a follow-
up measurement for static loads in the K-BNME must be carried out after the
measurements in the K-BNME, which is higher in the calibration hierarchy, or the K-NME.
The further investigations relate to the proof of suitability of the continuous calibration
process. First of all, the influence of the load speed up to the minimum load time with
continuous operation is to be determined. Then one or more transfer force transducers
are continuously calibrated with the K-BNME to be connected and the results are
compared with the results of a calibration of the KAtsd on a K-BNME or a K-NME that is
higher in the calibration hierarchy. If the results of the continuous calibration are within
the measurement uncertainty specified for the process, the proof of suitability is provided.

K-NME
PTB
K-BNME
DKD
reference force transducer transfer force transducer
Static calibration e.g. Static calibration Static calibration e.g. B.
B. according to DIN EN ISO 376 according to DIN EN ISO 376 z. B. according to DIN EN ISO 376
relief creep unloading creep unloading creep
influence of influence of
load speed up load speed up
sensitivity and reversal range sensitivity and reversal range
K-BNME
Static connection measurement
Continuous calibration of the transfer

Force transducer and comparison of
Results with those of the static
calibration
Figure 14: Feedback for the continuous calibration process


Gradual
calibration
National Standard
K-NME
reference standard
K-BNME
transfer standard reference standard

KA KA
thousand ref
Gradual calibration
according to DIN EN ISO 376 or similar.
Proof of suitability of the force standards continuous

for continuous calibration
calibration
modeling modeling
3rd order 3rd order
compensation function compensation function
Calibration in
continuous
transfer reference reference reference
Calibration device
calibration data calibration u. work data
with KA as reference
standard,
KA Ref as a calibration item
thousand
correction
Calibration Calibration data

detectable data tiered continuously continuous
calibration
suitability
process
Proof
the
of
systematic
deviations
comparison comparison
Random and undetectable systematic deviations
Rel. ext. Connection measurement uncertainty Wbmc, cont of

continuous calibration
Figure 15: Proof of suitability for the continuous calibration process

DKD-R 3-9
doi.org/10.7795/550.20180823J Side: 40 / 65
A.2 Requirements for reference and transfer force transducers
For reference and transfer force transducers, it is recommended that the following requirements be
met:
Reference force transducer Transfer force

K-BNME transducer
Type wbmc,cont Minimum
9 10
classification according to wref wtsd
EN ISO 376
A 0.10% ÿ W ÿ 0.25% 00 W ÿ 0.06 % W ÿ 0.04%
B 0.25%< W ÿ 0.5% 0.5 W ÿ 0.12 % W ÿ 0.08 %
C 0.5% < W 1 W ÿ 0.24% W ÿ 0.16%
Table 9: Recommendations for reference and transfer force transducers
Furthermore, the following properties must be observed:
Stable, non-time-critical measuring behavior (characteristic value, hysteresis, residual pre-

force signal) low creeping low adiabatic influence (thermoelastic creeping) moment-free force
introduction and discharge low bending moment interference behavior, possibly rotationally
symmetrical measuring behavior
Temperature compensation of the characteristic value TKC and the zero signal TK0
For calibrations in the direction of tension and compression with zero crossing, i.e. in an installation
situation, suitable force transducers must be used both as reference and transfer force transducers
for tension and compression. The load sense must be the same for all calibrations and traceability
measurements.
9 Limits of the relative expanded measurement uncertainty for different classes according to
EA-10/04 (formerly EAL-G22), Table 9.2. The contributions to the assessment of the measurement
uncertainty depend on the design of the reference or transfer force transducer 10. For follow-up
measurements using transfer force transducers, the same limits for the relative expanded
measurement uncertainty apply as under footnote 1, but reduced by the contribution from the
reversal error

transducers using the comparison Output: 09/2018
method https://doi.org/ Revision: 0
10.7795/550.20180823J Side: 41 / 65
A.3 Proof of suitability of the transfer standards before the first use
of continuous operation (proposal)
The suitability must be proven in addition to the previous static calibration procedure with additional tests
in the calibration hierarchically higher K-BNME or the K-NME.
A.3.1 Determination of the influence of the load speed on the measured values Two series of measurements
according to a) and b) in the same installation position are required to determine the influence of the load
speed on the measured values:
a) Rapid loading in steps of 20 … 25% up to the end of the measuring range with intermediate
unloading after each step (see Figure 16).
The measured value Sjump,i is determined after a short waiting time tW (eg 2 s, depending on the machine)
after the load has been applied. The prerequisite is that transient processes have faded away. The sum of
the load times of all stages should not be greater than twice the load time tW,kont aimed for in continuous
operation :
N
ÿ tÿ ÿ 2 t
W, i W, cont
i =1
sjump,i
t/s
Figure 16: Procedure for determining the influence of the load speed on the
readings
b) Staged loading procedure and measurement value determination

Sstat,i according to the specifications of DIN EN ISO 376 at the same load levels as in a).
The rel. Measured value difference determined: S S ÿ S ÿ = S

ÿ Sstat, i
i jump, i
rel, i
= (A.8)
i state
Si state
These values are taken into account in the following measurement uncertainty budget to quantify the
influence of "load speed on the measured values".

A.3.2 Determination of the influence of loading rate on reversal

span
Two series of measurements in the same installation position are required to determine the
influence of the load speed on the hysteresis. This measurement is only necessary if the
hysteresis is included in the calibration process.
a) Staged stress cycle according to p.41 a), but stressing and relieving. The sum of
the load times of all stages should not be greater than four times the load time
tW,kont aimed for in continuous operation :
N
ÿ tÿ ÿ 4
W, i
tW, cont
i =1
The relative Turnaround for each stage is given by:

S Sjump,on , i ÿ
jump, off, i
u
jump, i
= (A.9)
S
jump,on , i
shortest from the K-NME or

K-BNME realizable time
ujump,i
t/s
Figure 17: Procedure for determining the influence of the load speed on the
reversal margin
Note: Figure
16 and Figure 17 only serve to show the principle of the load profile. The real process
depends on the possible time management of the force calibration device.
b) Load cycle based on DIN EN ISO 376 with the same gradation as on page 41 under
a) (20%… 25% up to the end value of the measuring range), but with loading and unloading.
The relative Turnaround range for each stage is given by:
S stat,ab,Si ÿ
stat,on , i
u
i stat
= (A.10)
Sstat,on , i
The rel. Influence determined as reversal span difference:


ÿ=
u ui ÿ
u i stat (A.11)
jump, i
These values are taken into account in the following measurement uncertainty budget
to quantify the influence of "load speed on the hysteresis".
A.3.3 Creep behavior test A basic

distinction is made between loading and unloading creep (see Figure 18).
Both effects deliver approximately the same creepage values, taking into account the
measurement uncertainty sought here. It is therefore usually sufficient if only the
unloading creep is measured. The unloading creep can be determined on every K-BNME
- also in the comparison method . On the other hand, the exact determination of the load
creep requires a direct load device11 .
Notes: 1.
The unloading creep can be approximately estimated from the preload creep (cf. [30]).
2. To ensure that the load history of the force transducer has no influence on the creep
behavior, the transducer must be loaded at least once up to the nominal force. Then you
have to wait 10 minutes before starting the creep measurement.
t3 , S3
t2 , S2
S2
t 1, S1
Output
signal
S
t4 , S4
t 5, S5
t 6, S6
t
0
SF0
tv 5tv tF tW tB tF tW tB
time t
Figure 18: Creep behavior of a force transducer (asymptotically incoming creep

function), from VDI/VDE 2638, Figure 3, Section 3.4.3.1
11 By determining the loading and unloading creep, the estimation of the measurement
uncertainty can be made more precise.

doi.org/10.7795/550.20180823J Side: 44 / 65
t0, SF0 Time at the beginning of the change in force with the associated pre -force signal SF0
(Output signal before loading at time t0)
t2, S2 Point in time at the beginning of the evaluation time of the load creep with the
associated output signal S2 (output signal after the load at point in time t2 )
t5, S5 Point in time for evaluating the load zero drift and start of the evaluation time of
unloading creep with the associated output signal S5 (output signal after tW
, at time t5)
t6, S6 Time t6 of the end of the evaluation time of the unloading creep with associated
output signal S6
Typical times for forces up to 100 kN are:

Force change time: tF eg 2 s to time:
3 s Waiting
tW eg 2 s
Evaluation time: tB eg 10 min
For larger forces, the times depend on the loading devices.
The following are to be determined (see Figure 18):
= S5S
ÿ
6
unloading creep: i.ecr, E (A.12)
C noun
= S 6S
ÿ
f0
pre-force signal rest: i.e0,R (A.13)
C noun
with Cnom as the nominal characteristic
A.3.4 Determining the influence of the force introduction parts If

other force introduction parts are used in the connection measurement than in the calibration of the
transfer force transducer in the K-BNME or the K-NME, which is higher in the calibration hierarchy,
their influence on the KAtsd must be examined separately and a proportion in the measurement
uncertainty budget to be taken into account.
A.3.5 Determining the influence of temperature The

temperature coefficients TK0 and
andTKC of should
theforce
transfer reference
transducers with
comply
used
the limit values specified in Table 10. If no own
measurements are possible at different temperatures, the manufacturer's data sheet information
should be used as a basis.
DKD-R 3-9
10.7795/550.20180823J Side: 45 / 65
A.3.6 Recommended limits for time-dependent parameters and temperature behavior

ten
Reference force
wbmc,cont transfer force
criteria transducer or
Type transducer
K-BNME
Stable, non-time critical ÿS rel.i A

ÿ 0.3% Wref ÿ 0.3% WTsd
measurement behavior b/c
A ÿ 0.015 % ÿ ÿ 0.010 % ÿ
creep influence dcr,E
b/c 0.020 % ÿ 0.015 % ÿ
i.e
A 0.012 % ÿ 0.008 % ÿ
pre-force signal remnant 0,R
b/c 0.025 % ÿ 0.016 % ÿ
reversal margins ÿu A 0.025 % ÿ 0.015 % ÿ
i
influence b/c 0.050 % ÿ 0.025 % ÿ
A 0.0010%/K ÿ 0.0008%/K ÿ
TK0
b/c 0.0020%/K ÿ 0.0012%/K ÿ
temperature behavior A
TKC 0.0015%/K ÿ 0.0010%/K ÿ
b/c 0.0030%/K 0.0015%/K
Table 10: Limit values for time-dependent parameters and temperature behavior of the KAref and
katsd
A.3.7 Monitoring of continuous operation To monitor continuous

operation, the following measures in particular are necessary with regard to the reference force transducer:
Monitoring of the residual pre-force signal

Monitoring of the sensor temperature
Self-monitoring: Measurements with transfer force transducers in accordance with the QMH
A.4 Determination of the smallest measurement uncertainty that can be

specified for a K-BNME with a reference force transducer system
in continuous operation (suggestion)
The method for determining the smallest measurement uncertainty that can be specified for continuous
operation is based on that for static operation [42].
A.4.1 Influencing variables

In the case of continuous operation, at least the following influencing variables must be taken into account
in addition to those listed in the publication mentioned above. These are shown in the cause and effect
diagram in Figure 19.
DKD-R 3-9
10.7795/550.20180823J Side: 46 / 65
Realization (K-NME) reference force transducer transfer force transducer
calibration certificate calibration certificate
temperature temperature
long-term instability long-term instability
power
reversal margin
Std Dev. (K-NME) difference reading
difference
Std Dev. (K-BNME) Suitability (reference and transfer force transducer)
reversal
span (cont.)
system. dev. (stat.) system.-
dev. (cont.)
Std Dev. (cont.)
reversal margin Std Dev. (stat.)
(K-BNME)
Verification (K-BNME)
connection, stat. connection, cont.
Figure 19: Factors influencing the determination of the smallest measurement uncertainty that can be specified
with continuous operation
A.4.2 Model
The model of force generation by a K-BNME with a reference force transducer system is formulated as a
linear product model using relative measurement uncertainties.
In general, the following applies:
N
Sizes: =ÿ
Y q X pÿ
p
ii
; i
:= ± 1 (A.14)
i=1
N
estimated values: =ÿ
yq x ÿ i
pi
; pi = ± 1 (A.15)
i
In the detailed representation - conditional equation with the "best" estimated values - this becomes the
expression:
3 3
F
bmc, cont
=ÿ
FKRef i ,stat ÿÿ ÿ
K
i , cont
i =1 i=1
3 3
(A.16)
With ÿ KK
i ,stat
=
drift_ref
ÿ
K Kÿ ÿ
connection, stat
and ÿ =
KKK K i ,cont
Sÿÿu
ÿ ÿ
ver
i =1 i=1

doi.org/10.7795/550.20180823J Side: 47 / 65
12
result size; force acting on a calibration item during continuous operation
fbmc,cont
13 14
FRef Set force (reference force transducer)
KDrift_Ref correction factor rel. Long-term drift of the reference force transducer
Kÿ correction factor rel. Change in ambient temperature
Kconnection,stat correction factor Connection measurement with transfer transducers
KÿS correction factor rel. Measured value difference (in K-NME)
cÿu correction factor rel. Reversal margin difference (in K-NME)
Kver Correction factor verification of continuous calibration
Correction Factors:
ÿ ÿÿ Fi
Ki ÿ
= ÿ+ ÿ ÿ 1 (A.17)
fi
With
ÿFi unknown measurement deviations and
E[ÿ ] 0 Fi = expected value
Due to the selected, linear product model and the consideration of relative
measurement uncertainties, the sensitivity coefficients are always equal to ±1*i (=c±1 ).
Based on the determination of the smallest specifiable measurement uncertainty in static

operation, additional contributions that are necessary for estimating the smallest specifiable
measurement uncertainty in continuous operation must be taken into account with continuous
calibration.
This results in the standard measurement uncertainty:
( bmc, cont) = w2 F( wref K) + 2

( 2 2 2 2
( ÿ S ) + ++( 2 (A.18)
wF drift_ref ) ( ÿ )++
wKwK ( )
connection, stat
wKwKwK ÿ and
) ( ver )
12
result size

10.7795/550.20180823J Side: 48 / 65
The associated distribution functions and rel. Variances can be found in Table 11
men:
2
FRef normal FRef
()w
2a
drift_ref
KDrift_Ref rectangle 2 w( K drift_ref )=
3
2 ÿa
Kÿ rectangle (
wK
2
ÿ )=
3
ÿ 2
ÿ ÿF ÿ
stat
Kconnection, stat rectangle Uÿÿ f KNME
2wK
( )=
connection, stat
ÿÿ
2 2a
ÿS
KÿS rectangle w K( ÿS )=
3
2 2a
ÿu
cÿu rectangle (
wK ÿ and
)=
3
2 2a
rectangle ver
Kver wK( = ver )
3
Table 11: Distribution functions and variances for continuous loading for the model for Fbmc,cont
DKD-R 3-9
10.7795/550.20180823J Side: 49 / 65
A.4.2.1 Uncertainty analysis The

No.
*
Xi xi a w(xi) ci wi(y) 2 w (y)
2
1 x 1 ( cont
w f F bmc, ) ( cont
w fF bmc, )
FRef 2 w F( Ref ) 3 ref ref
2
2 KDrift_Ref 1 a drift_ref 1 wdrift_ref
F wdrift_ref
F bmc, ( cont )
w K( Drift_Ref ) bmc,cont ( )
2
3 Kÿ 1 aÿ 3 w K( ) 3 w ÿ 1wF ÿ
( bmc, account)
wÿ F( bmc, cont
)
4 1 a connection, stat 1 w f 2f ( F bmc, cont)
Kconnection, stat K( Connection,stat ) An...,stat bmc,cont () An...,stat
2
5 KÿS 1 S aÿ 3 w K( ÿS ) 1 wÿS
F ( bmc,cont ) w F ÿS
bmc,cont ( )
2
6 cÿu 1 aÿ and
3 w K( ÿu ) 1 wÿu
F ( bmc,cont ) w F ÿu
bmc,cont ( )
7
Kver
1 a
ver 3 w K( ver ) 1 w ver
F bmc,cont ( ) ver F(
2w bmc, cont)
N
y rel. standard measurement uncertaintyw F wi2F(
fbmc,cont
w
( bmc, cont )=ÿ bmc, cont )
i= 1
y rel. expanded measurement

fbmc,cont WF( bmc,cont )kw
= Fÿbmc,
( cont )
uncertainty W
Indication of the complete measurement result Fbmc,kont = ±y W F( bmc,kont)
Table 12: Uncertainty analysis for the model for Fbmc,kont
DKD-R 3-9
Side: 50/65
A.4.2.2: Measurement Uncertainty Budget

power verification
reference recording Ref. pickup drift ambient temperature connection, measured value difference Reversal margin continuous
more static renz difference
calibration
kN #
at least
Max.
rel. Standard measurement uncertainty wi(y) rel. Standard measurement

rel. Expanded measurement
(combined measurement uncertainty contributions uncertainty w(y)
uncertainty W(y) (k=2)
power associated with the result due to influencing variables)
Continuous operation
reference transducer Static operation In total
#
kN
at least
Max.
Table 13: Measurement uncertainty budget for determining the smallest measurement uncertainty that can be specified in continuous operation

doi.org/10.7795/550.20180823J Side: 51/65
A.4.3 Correction factor Kconnection, instead of the static connection measurement with
transfer force transducers
Model error :
The influence due to the static connection measurement with transfer force transducers
can be formulated in a general representation as a sum/difference model:
Sizes: Y=
p ÿX p ;i ÿ
i i
1=±
(A.19)
i=1
estimated values:
=
; ypxp =±1
(A.20)
ÿ
ÿ i i i
i =1
In the detailed representation - equation for determining the measurement error with the
ÿ =f stat
F KBNME
ÿ
f KNME
+ ÿ ÿf i ,stat
i= 1
3
(A.21)
with ÿ ÿ f
i ,stat
= ÿ Fÿ
realization
F+ drift_thsd
+ ÿF
HysKBNME
i=1
15
ÿFstat result size; Deviation of the mean force values under static load
16 17
F KBNME Force display mean value of the transfer force transducer in the K-BNME
F KNME Force display mean value of the transfer force transducer in the K-
ÿF NME unknown error
realization
Force representation in the NMI (during
ÿF implementation) unknown measurement deviation
drift_thsd
Long-term drift of the transfer force
transducer unknown measurement error
ÿF
HysKBNME Reversal range (hysteresis) of the K-BNME, determined taking into
account the reversal range of the transfer force transducer in the K-NME
15
result size

doi.org/10.7795/550.20180823J Side: 52 / 65
Unknown errors: ÿFi unknown

errors
,stat and
E Fiÿÿ ÿ,stat ÿ = ÿ 0 expected value
Due to the selected sum/difference model with the coefficients equal to 1, the sensitivity
coefficients are always equal to ±1 ( ci = ±1).
(F
u) ÿ&stat
=F
2
( KBNME uF
2
(+ KNME 2
ÿuF(+ ) 2
ÿuF(+ ) 2
ÿuF + (A.22)
) ) realization drift_thsd ( HysKBNME )
The associated distribution functions and variances can be found in Table 14:
random variable Variance u²(xi)
n 2
( ÿ ( xx ÿ
KBNME i )
F KBNME normal 2 i= 1
uF ( KBNME )=
n ÿÿ
1 ) n
n 2
(ÿ xxi ÿ
KNME )
F KNME normal 2 i=1
u F( KNME )=
(n ÿÿ
1 n
ÿF u 2Fÿ
realization normal ( realization ))
2a
ÿFdrift_thsd rectangle ÿ 2u(F )= drift_thsd
drift_thsd
3
2a
ÿF rectangle ÿ 2u(F )= HysKBNME
HysKBNME
HysKBNME 3
Table 14: Distribution functions and variances for the model Deviation (static
connection measurement)

doi.org/10.7795/550.20180823J Side: 53 / 65
A.4.3.1 Uncertainty Analysis The

No.
Xi xi a u(xi) ci ui(y) 2ui _ (y)
1 ×1 2 s 1 and
( Fÿ ) ( Fÿ )
F KBNME f
KBNM E stat
2 &f
KBNM E stat
n
2 2 s -1 and
( Fÿ ) ( Fÿ )
F KNME x2 KN ME F stat
2 &f
KN ME stat
n
3 ÿF 0 2 u (realization) 1 and f (ÿ )
stat 2& (Fÿ stat )
realization realization realization
4 ÿF 0 a 3 u (Drift_Tsd) 1 and f (ÿ )
stat 2& ( Fÿ stat )
drift_thsd drift_thsd drift_thsd drift_thsd
5 ÿF 0 a 3 u(HysKBNME) 1 and f (ÿ )
stat 2& (Fÿ stat )
HysKBNME HysKBNME HysKBNME HysKBNME
F) = ÿ ÿ f( iand
2
ÿFstat y standard measurement uncertainty ( ÿ u stat stat )
i=1
ÿFstat y Expanded measurement uncertainty U UF ku F ( (ÿstat= )ÿ ÿ stat )
Indication of the complete measurement result ( ÿstat= ± F y U ÿF s at t )
Table 15: Uncertainty analysis for the measurement error model (static connection meas
sung)
DKD-R 3-9
Side: 54/65
A.4.3.2 Measurement Uncertainty Budget
Standard measurement uncertainty ui(y)

(measurement uncertainty contributions associated with the result due to influencing standard advanced
variables) measurement measurement
power uncertainty uncertainty
ÿF ÿF ÿF u(y) U(y) (k=2)

FKBNME FKNME realization drift_thsd HysKBNME
kN #
at least
Max.
Table 16: Measurement uncertainty budget for the static measurement deviation (static connection measurement with transfer force transducers)
DKD-R 3-9

10.7795/550.20180823J Side: 55 / 65
The determined deviation of the mean force values is to be applied as a correction.

However, this is not done in practice; the deviation ÿFstat and the associated expanded
'
u
measurement uncertainty UF (ÿ stat) are combined to form the uncertainty interval ÿF . stat
A rectangular distribution is assigned to the uncertainty interval. For use in the higher-
level model, this uncertainty interval is to be related to the mean force value F KNME .
ÿ
u ÿF = ÿ +f ÿ FU
stat ( stat
) (A.23)
stat
ÿ uÿf ÿ
stat
K connection, stat = ÿ ÿ 1 (A.24)
ÿÿ f KNME
ÿÿÿÿ
A.4.4 Correction factor Kver of the verification of the continuous method (static and
continuous measurement with transfer force transducers)
model error
The influence of the continuous process, quantified by static and continuous
Measurement with transfer force transducers in the K-BNME can be represented in general as
sum/difference model can be formulated:
N
Sizes: Y =p ÿX p ;i ÿ
i i 1=± (A.25)
i=1
N
estimated values: y = ÿ ; pxp
i i
ÿ
i 1=± (A.26)
i =1
In the detailed representation - equation for determining the measurement error with the
ÿ FFF
= ÿ +Fÿ (A.27)
ver cont stat Hys, cont
18
result size; Deviation of the mean force values between continuous
ÿFver
and static load
19 20
Force reading mean value of the transfer force
F cont
transducer with continuous load
f Force display mean value of the transfer force
stat
transducer with static load unknown measurement
deviation
ÿF Reversal range (hysteresis) with continuous load, determined
Hys, cont
taking into account the reversal range of the transfer force
transducer with static load
18
result size
20 N input values for determining the measurement uncertainty N'<N
DKD-R 3-9

10.7795/550.20180823J Side: 56 / 65
Unknown errors: ÿFi unknown

errors
,cont and
ÿ=ÿ
E Fiÿÿ ÿ ,cont 0 expected value
Due to the selected sum/difference model with the coefficients equal to 1, the sensitivity
coefficients are always equal to ±1 ( ci = ±1).
2 2 2
(ÿf = ver fu F(u
)+ ( (F
u +ÿu (A.28)
) cont stat
) Hys, cont )
The associated distribution functions and variances can be found in Table 14:
random variable Variance u²(xi)
n 2
normal
( ÿ xx i cont ÿ
)
F con t i=1
(
2 u F cont = )
(n ÿÿ
1 ) n
n 2
normal
( ÿ xx i ÿ
stat )
f 2 i= 1
stat
uF ( stat )=
(n ÿÿ
1 ) n
2a
ÿF rectangle 2
)ÿ u F =
Hys, cont
( Hys,kont
Hys, cont
3
Table 17: Distribution functions and variances for the model Deviation (verification of the
continuous method)
DKD-R 3-9

10.7795/550.20180823J Side: 57 / 65
A.4.4.1 Uncertainty analysis The

No.
Xi xi a u(xi) ci ui(y) 2ui _ (y)
1 ×1 2 s 1 and
f (F ÿ v he ) &
2
f (F ÿ v he )
fcont cont cont
n
2 2 s -1 u Fÿ( ) u 2Fÿ( )
Fstat x2 n
F stat ver F stat ver
5 ÿF 0 a 1 f (Fÿ )
3 u (H ys,kon t) 2&
and
Hys, cont ver (ÿ ) Hys, cont ver
Hys, cont Hys, cont
N
2
y standard measurement uncertainty
ÿFver u (F ) ÿver= ÿ ÿ f( i and ver )
i= 1
ÿFver y Expanded measurement uncertainty U UF ku F ( (ÿver=) ÿ ÿ ver )
Indication of the complete measurement result ( ÿver= ± F y U Fÿver )

Table 18: Uncertainty analysis for the measurement deviation model (verification of the continuous method)
exposure level. The following measurement uncertainty budget (Table 19) provides a clear overview:
DKD-R 3-9
Continuous calibration of force transducers using the 1
Output:
comparison method https://doi.org/ Revision: 0
10.7795/550.20180823J Side: 58/65
Standard measurement uncertainty ui(y)

(measurement uncertainty contributions associated
standard measurement advanced
with the result due to influencing variables)
uncertainty measurement uncertainty
power
u(y) U(y) (k=2)

ÿF
fcont Fstat Hys, cont
kN #
at least
Max.
Table 19: Measurement uncertainty budget for the measurement deviation (verification of the continuous method, measurements with transfer force appr
takers)
DKD-R 3-9

10.7795/550.20180823J Side: 59 / 65
The determined deviation of the mean force values is to be applied as a correction.

However, this is not done in practice; the deviation ÿFver and the associated expanded
´
measurement uncertainty UF ( ÿver ) are combined to form the uncertainty
ver
interval UÿF .
A rectangular distribution is assigned to the uncertainty interval. For use in the higher-
level model, this uncertainty interval must be related to the mean force value Fstat .
ÿ
U ÿF = ÿ +FU
ÿ F ( ) (A.29)
ver ver ver
ÿ
u ÿ f ver
K =ÿÿ1
(A.30)
ver
ÿÿ f stat
ÿÿÿÿÿ
Appendix
B B.1 Determination of an optimal polynomial degree for reduction
tion of the influence of random deviations
With continuous calibration, the influence of random deviations cannot be reduced by
averaging the measured values, as is usual with stepwise calibration. If necessary, other
methods must therefore be used, such as digital filtering or compensation functions
(model-based reconstruction).
The following explanations go into the determination of compensation functions and
their optimal degree n . These are based on system-technical considerations using the
example of a piezoelectric transducer.
For the model-based reconstruction of the calibration item output signal, the calibration
model f must be set up and its parameters estimated. This is done separately for the
increasing and decreasing load series.
SKG
f F = ( Inref
general, the following applies to the calibration )
model: where SKG is the output signal of the calibration item, which is reconstructed
using the calibration model.
Polynomials of degree n are usually used as the analytical form of the model :
n
SKG = ÿ aÿ i f
i
ref
(B.1)
i =1
The parameters i are estimated a using the (Gaussian) method of least squares. The force
Fref,kont of the reference force transducer is considered as the independent variable
and the output signal SKG,kont of the calibration item as the dependent variable.
When choosing the polynomial degree n , it must be ensured that there is a sufficiently
large number of degrees of freedom nF : nF = N – n. N is the number of measured values
in the data series. If possible, this should have 100 or more values per load series and
direction in order to ensure a well-founded statistical evaluation of the data with regard
to randomness, freedom from outliers and normal distribution.
If the number of degrees of freedom is too small, the model function will be influenced
by random deviations because the confidence intervals of the estimated parameters
become too large. Polynomials with degrees in the range 3 ÿ n ÿ 6 are usually sufficient for the syste
DKD-R 3-9

10.7795/550.20180823J Side: 60 / 65
Depict the behavior of the transducer in the model. To verify the method, it is
important to check the model. For this purpose, different degrees of the model
polynomial are first assessed by comparing the associated sums of the squared
deviations. If the sum of the squares of the deviations no longer decreases significantly
when the degree n is increased, an increase in n no longer makes sense. For the example
shown, a 5th degree polynomial is optimal.
2000
[pC2 ]
2
1500
ÿr
1000
500
0
3 4 5 6
Degree of the polynomial n [-]
Figure 20: Representation of the sum of the squared deviations as a function of the degree of the polynomial.
Since the least squares method is strongly influenced by outliers and is therefore not
very robust, outliers must be recognized and eliminated. Various methods can be
used to detect outliers. When checking for normal distribution using a quantile-
quantile plot, outliers are visible as points far from the straight line. For larger data
sets, the simplest method is to exclude pairs of values that result in residuals that are
too large as outliers. –3sr > r > 3sr is often used as an exclusion criterion . After
excluding pairs of values that led to outliers, the coefficients of the polynomial must
be recalculated, and an iterative check for the absence of outliers may be necessary.
The optimally adapted model polynomial can be subjected to further testing. Since
the least squares method is only optimal for normally distributed deviations (residuals
r) between model and data, the residuals are checked for normal distribution. This is
preferably done using a "normal distribution paper" or the so-called quantile-quantile
plot (QQ plot). If the residuals lie largely on a straight line, the residuals are randomly
distributed and the model is therefore optimally chosen (Figure 21).
0.99
0.98
0.95
0.90
0.75
0.50
probability 0.25
0.10
0.05
0.02
0.01
0 1 2 3 4 5 6
Residuals r [pC]
Figure 21: Visual check of the residuals for normal distribution using the quantile quantile
plots
DKD-R 3-9

10.7795/550.20180823J Side: 61 / 65
With the help of the optimized and tested calibration model, the output signals SKG of
the calibration object can now be reconstructed. The standard deviation of the residuals
sr of the optimal model is regarded as a random deviation and thus contributes to the
measurement uncertainty.
B.2 Example of the effect of the approximation with polynomials The following
figures show the effect of the approximation of directly recorded pairs of values using
polynomials. Deviation diagrams are shown for illustration, in which the transmission
coefficient as the slope of a straight line through zero was used as the result of the
calibration. The straight line was determined by minimizing the absolute deviation
amounts (minimum method with Chebyshev approximation). The deviations were
determined for each recorded pair of values (Fref,kont; SKG,kont) from a data set with
90 measuring points each for the increasing and decreasing load cycle up to 60 kN.
Figure 22 shows the deviation diagram that results from the evaluation using directly
recorded pairs of values (Fref,kont; SKG,kont) .
0.1
0.08 directly recorded pairs of values - best straight line
0.06
0.04
0.02
%deviation
Relative
KG,kont
Straight
related
FSO
to
in
-S S
-0.02
-0.04
-0.06
-0.08
-0.1
0 10000 20000 30000 40000 50000 60000
Reference force F in N ref,cont
Figure 22: Deviation diagram when evaluating directly recorded value pairs Fref,cont
and SKG,kont
It can be seen that the directly recorded pairs of values contain random deviations of
the order of ±100 ppm, which can be traced back to signal disturbances such as noise.
If linear interpolation (according to paragraph 7, method 1a) is used, these random
deviations are transferred to the pairs of values at the desired interpolation points Fref
and SKG. In order to reduce the random deviations, according to paragraph 7,
procedure 1b, a compensation function of a higher degree can be placed over the
directly recorded pairs of values, with the ascending and descending load cycle being treated separa
DKD-R 3-9

10.7795/550.20180823J Side: 62 / 65
the compensation function determines the value pairs (Fref; SKG) at the desired interpolation
points as a reconstructed output signal of the calibration item. The reconstructed output
signal at the interpolation points is used for evaluation via polynomial formation in order to
determine the straight line using the minimum method with Chebyshev approximation. The
relative deviation between the reconstructed pairs of values and the best straight line is
shown in Figure 23. The basic course of deviation between the pairs of values and the best
straight line has not changed over the calibration range of 60 kN.
0.1
Values of the output signal with interpolation -
0.08 Best straight
0.06 Reconstructed values of the output signal per

Polynomial - best straight line
0.04
0.02
Straight
related
FSO
KG
%to
in
-S
0
-0.02
Deviation
Rel.
S
-0.04
-0.06
-0.08
-0.1
0 10000 20000 30000 40000 50000 60000
Reference force F ref in N
Figure 23: Deviation diagram when evaluating certain pairs of values at the interpolation
points Fref and SKG by forming polynomials. The random deviations between
the compensation function and the directly recorded pairs of values become a result of the
approximation as the residual standard deviation sr and are included as a contribution to the
estimation of the measurement uncertainty. They no longer have a direct effect on the pairs
of values at the desired support points Fref and SKG . For comparison, the residual standard
deviations with sr = ±56 ppm at each reference point were entered as a scatter range in
Figure 23. In addition, the output signal is mapped at the desired interpolation points, which
results from a linear interpolation between the directly recorded pairs of values according
to method 1a. It becomes clear that this output signal depends on the random fluctuations
in the measured value acquisition.
nomenclature
[N]
FRef, cont directly recorded force values during the continuous process
FRef Force at a reference

SKG, cont [N] [pc] point in the case of the continuous method directly recorded output
signals of the calibration object reconstructed value of the KG
SKG [pc] output signal
sr [pc] Standard deviation of the residuals of the calibration model
DKD-R 3-9

10.7795/550.20180823J Side: 63 / 65
N [-] Number of measured values of the data series

ai [-] Parameters of the model function
nF [-] number of degrees of freedom
n [-] degree of a polynomial
DKD-R 3-9

10.7795/550.20180823J Side: 64 / 65
Appendix C
C.1 Display Deviation As an

alternative to the evaluation models described, the "Display Deviation" model can be
used. It is assumed that the previously determined values of an equalization function
represent display values.
N
ÿ=
f F ÿ f+ÿ ÿf (C.1)
compensation normal i
i=1
Here, measurement uncertainties are used in the unit of force, ie in kN. This deviates
from the existing standards and regulations for the quantity of force, in which the
characteristic values characterizing the transducer are regarded as relative, ie related
dimensionless, values (analogous to EA-10/14: EA Guidelines on the Calibration of
Static Torque Measuring Devices, June 2000).
Editor:
Physical-Technical Federal Institute

German calibration service
Federal Avenue 100
38116 Brunswick
www.dkd.eu
www.ptb.de

DKD-R - 3-9 - 2018 (1) English

Uploaded by

Copyright:

Available Formats

DKD-R - 3-9 - 2018 (1) English

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DKD-R - 3-9 - 2018 (1) English

Uploaded by

Copyright:

Available Formats

Machine Translated by Google

policy Continuous calibration of

German Calibration Service (DKD)

Calibration laboratories from industrial companies, research institutes, technical

Suggested citation for citing the

Published by the Physikalisch-Technische Bundesanstalt (PTB) for the German Calibration

Purpose and Scope 6

symbols and designations 7

5 6 6.1 6.1.1 Determination of design-specific measurement uncertainty 10

contributions 6.1.2 Temperature compensation 6.2 Calibration process 6.2.1 10

evaluation 7 8 Measurement uncertainty 8.1 Influencing variables 8.2 Model of the 10

evaluation "Transmission coefficient" model 8.2.1 8.2.2 "Force" model (polynomial 10

coefficients of a 3rd degree adjustment, DIN EN ISO 376) 11

2 Purpose and Scope

3 Symbols and designations The symbols

between fast and static loading unloading

TKC Temperature coefficient of the characteristic value

Wbmc can be specified for static calibration Smallest measurement

Table 1: Symbols and designations

operating instructions) are available

Functional tests include, for example:

4.2 Clamping parts, built-in parts

4.3 Signal-carrying components in the K-BNME

6 Performing the calibration

6.1 Preparations for calibration

6.1.2 Temperature compensation

6.2 calibration process

6.2.1 Preloads The 1st

6.2.2 Time management based on DIN EN ISO 376 Figure 1

6.2.3 Time management for shortened calibration

Figure 1: Continuous calibration process based on DIN EN ISO 376

6.2.4 Acquisition of the measured

The evaluation includes the following steps (see Figure 3):

2. Provision of measurement uncertainty contributions (e.g. repeatability, reproducibility, hysteresis)

When using a 1st degree compensation:

5. Representation of the transmission coefficient (single-value specification) with

6. Specification of the complete measurement result consisting of transmission

Installation position 240°

1) Determination of pairs of values at interpolation points by: b)

Free, cont Polynomial, ascending,

support points dividing x° Polynomial,

3) From the pairs of values at the interpolation points, determination of a

4) Set up the table of values

6) full Measurement result: Transmission 6) Table of values +

Figure 3: Process of an evaluation

8th measurement uncertainty

temperature temperature reversal margin

Comparison precision (rotation)

long-term instability long-term instability repeatability

indicator customizer calibration item

Selection of the appropriate measurement method

Other sources of measurement errors are described in Table 2 below:

Measuring device Calibration interval expired,

Table 2: Sources of measurement deviations and measurement uncertainties [19]

8.2 model of the evaluation

In general, the following applies:

8.2.1 "Transmission coefficient" model In the

ÿx i unknown measurement deviations and

8.2.1.1 Measurement uncertainty